TL;DR
Get business pricing on networking and server gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Oracle announced a multivendor expansion of Oracle Cloud Infrastructure, with Nvidia, AMD and Intel systems and plans to add Google TPUs and Cerebras accelerators. The deployments span different schedules: Nvidia Vera Rubin availability is described as starting now, Google TPU capacity is planned from 2027, and Cerebras CS3 deployments are also planned for 2027. Capacity, pricing and availability details vary and are not fully specified.
Oracle announced a multivendor expansion of its Oracle Cloud Infrastructure (OCI) at AI World 2026, adding or planning systems from Nvidia, AMD, Intel, Google, Cerebras and SambaNova to serve AI training, inference and enterprise computing. The announcement brings together deployments at different stages: Oracle says Nvidia Vera Rubin systems are becoming generally available, while Google TPU and Cerebras deployments are planned for 2027.
Oracle said the expansion is part of a longer-term buildout of OCI, which it described as spanning more than 200 cloud services. The company also reported that it delivered an additional 850 megawatts of data-center capacity during its first-quarter fiscal 2027 earnings period. That figure describes capacity delivered in that reporting period; Oracle did not provide a comparable baseline or a breakdown of how much is dedicated to the newly announced systems.
For Nvidia, Oracle said OCI will deploy Vera Rubin NVL72 systems at multi-gigawatt scale, alongside Vera CPU servers and NVL8 HGX systems. The announcement says general availability for the Vera Rubin deployment starts now. Oracle’s described Superclusters combine the NVL72 racks with Oracle Acceleron networking, Nvidia Spectrum-X Ethernet and Spectrum-6 switches, and BlueField-4 data processing units. Each NVL72 rack connects 72 GPUs using Nvidia NVLink; Oracle says its wider network can scale across hundreds of thousands of GPUs.
Oracle also outlined processor plans with AMD and Intel. Sixth-generation AMD EPYC processors will power Compute E7 instances, which Oracle says are designed for uses such as agent orchestration, concurrent agent sandboxes, retrieval services, in-memory caches and enterprise application tiers. AMD’s Helios AI rack is planned to use Instinct MI455X GPUs. Separately, a multiyear Oracle-Intel collaboration is intended to expand Xeon-based OCI computing, including planned X14 offerings using Intel’s next-generation Diamond Rapids processors.
The announcement positions OCI as a cloud platform offering customers more than one accelerator and processor path, rather than tying AI workloads to a single hardware supplier. Different systems can be suited to different tasks: large GPU clusters for training and inference, general-purpose CPUs for application and agent workloads, and specialized accelerators for particular serving requirements. That flexibility may matter to businesses trying to match performance, availability and cost to a workload, though Oracle did not publish comparative pricing or performance results across the announced platforms.
The plans also put networking and resilience at the center of the expansion. Large AI jobs depend on many processors exchanging data, so Oracle’s emphasis on Acceleron fabrics, multiple network paths and failover speaks to how it intends to connect and operate the hardware as deployments grow. The practical impact will depend on how quickly capacity becomes available, what customers can access in each region, and how the systems perform under real workloads.
Adding Google TPUs would be a notable change in the cloud market if the stated plan proceeds: Oracle says it will be the first hyperscaler outside Google to deploy the accelerators. That could give some customers another option for workloads designed for TPU architecture. However, the plan begins in 2027, and the announcement does not specify initial locations, customer access arrangements or the scale available at launch.
Products Worth Considering
Jackery HomePower 3000 Portable Power Station with Two 200W Solar Panels
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Black
As an affiliate, we earn on qualifying purchases.
OCI’s Multi-Vendor Infrastructure Plan
Oracle says its infrastructure already uses a range of Nvidia GPUs, including P100, V100, A100, H100, H200, GB200 and GB300 products, and AMD MI300X and MI355X accelerators. The new announcements extend that existing mix with future Nvidia Vera Rubin systems and additional AMD and Intel platforms. These are company-provided descriptions of OCI deployments; the source material does not include an independent inventory of installed capacity.
Oracle’s Acceleron networking effort is presented as the connective layer for these systems. The company says it combines dedicated network fabrics, converged network interface cards, zero-trust routing and multiple network paths. Oracle also announced extensions for front-end remote direct memory access connectivity and additional NVMe block-storage attachment options across bare-metal and virtual-machine environments, as well as a Collective Communications Library for distributed AI workloads.
The partner announcements do not all describe the same stage of availability. Oracle characterizes Nvidia Vera Rubin availability as starting now, while the Intel collaboration covers current and future Xeon platforms and the X14 Diamond Rapids offering is planned. Google TPU deployments and the first Cerebras CS3 deployments are both scheduled to begin in 2027. Oracle named SambaNova as another inference partner, but the supplied announcement details do not give a product, launch date or deployment scale for that work.
“AI inference demands choice. As agents take on more complex work, customers demand compute that matches each application’s performance needs.”
— Mahesh Thiagarajan, president of Oracle Cloud Infrastructure
Availability, Scale and Performance Details
Several commercial and operational details remain unspecified. Oracle did not provide customer pricing, regional availability, capacity by system, or a schedule for all Nvidia, AMD and Intel offerings. The phrase multi-gigawatt scale applies to the planned Nvidia Vera Rubin deployment, but the announcement does not state a delivery timetable or explain how much capacity will be available at each stage.
Oracle cited expected Google TPU benefits, including more than 50% fewer network hops and up to 80% better performance per dollar for low-latency serving. These are claims in the supplied announcement; it does not provide the comparison setup, workload, measurement method or baseline needed to evaluate them independently. Oracle also said TPU networking can sustain more than 97% goodput across thousands of chips during large-scale training runs, but did not provide supporting test details in the material provided.
The schedule and scope for SambaNova are unclear, as are the precise launch arrangements for Google TPUs and Cerebras systems. Oracle’s statements do not establish that every named platform is currently available to customers, and no independent confirmation or customer deployment data is included in the source material.
Deployment Schedules and Customer Access
The next developments to watch are customer availability and rollout details for the Nvidia, AMD and Intel systems, including the timing and regional footprint of OCI’s Vera Rubin capacity and AMD-powered E7 instances. Oracle’s X14 platform depends on the planned Intel Diamond Rapids generation, so further information from Oracle and Intel should clarify its schedule and specifications.
For the later-stage announcements, Oracle says Google TPU deployments are due to start in 2027 and ramp toward gigawatt scale over multiple years; the first Cerebras CS3 deployments are also planned for 2027. Oracle has not supplied a comparable timeline for SambaNova. More detail on launch regions, customer access, prices and workload performance will be needed to establish how the expanded portfolio works in practice.
Key Questions
What did Oracle announce?
Oracle outlined a multivendor expansion of OCI involving Nvidia, AMD and Intel systems, with plans to add Google TPUs and Cerebras accelerators. SambaNova was also named as an inference partner, but the supplied details do not specify its deployment plan.
When will the new systems be available?
Oracle says Nvidia Vera Rubin availability starts now. Google TPU and Cerebras CS3 deployments are planned to begin in 2027. The announcement gives no full availability schedule for every AMD, Intel or SambaNova offering.
What are AMD’s Compute E7 instances intended for?
Oracle says the instances, powered by sixth-generation AMD EPYC processors, are intended for agent orchestration, concurrent agent sandboxes, retrieval services, in-memory caches and web, API and enterprise application workloads.
Will Oracle be the first cloud provider outside Google to offer TPUs?
Oracle says it will be the first hyperscaler outside Google to deploy TPUs, with deployment planned from 2027. The supplied announcement does not give customer-access details or independent confirmation of the claim.
Did Oracle publish prices or independent performance comparisons?
No. The supplied material does not include pricing or independent comparisons. Performance figures for Google TPUs, including the stated performance-per-dollar improvement, are claims attributed to the announcement without test methods or baselines in the source.
Source: Wccftech
